import a Keras pre-trained network into Matlab

I'm having issues to import a pre-trained network I've trained in Python into Matlab.

- I used the function of: importKerasLayers ('ImportWeights',true) to import architecture & Weights

- I have 6 PlaceHolderLayers which are not supported by matlab - in 5 of them I can use transposedConv2dLayer custom layer to replace them

However, I'm not sure how do I import the Weights of those layers into my custom ones. I need them in order to use "assembleNetwork" function that will switch the imported network to a SeriesNetwork object that will be ready to use for prediction.

I followed this flow of work:

https://www.mathworks.com/help/deeplearning/ref/assemblenetwork.html

Answer 1

[Sivylla Paraskevopoulou]

Use the importTensorFlowNetwork function. importTensorFlowNetwork tries to generate a custom layer when you import a custom TensorFlow layer or when the software cannot convert a TensorFlow layer into an equivalent built-in MATLAB layer.

Note that you must first convert your TensorFlow model to the SavedModel format.

Answer 2

[Asvin Kumar]

You can use the 'Weights' property for each PlaceholderLayer.

placeholderLayers = findPlaceholderLayers(lgraph);
placeholderLayers(i).Weights

On the other hand, you could explore other parameters involved by checking the data members of the placeholderLayers.KerasConfiguration structure as in this example.

In general, a couple of details in this section might be useful to you.